opencodeiiita/medical-image-classification

In this open source project you will build a model which classify medical x-rays as Normal and Infected category.

Medical-Image-Classification

In recent years, the intersection of medical imaging and deep learning has witnessed unprecedented advancements, revolutionizing the landscape of healthcare. One notable application that has gained substantial attention is medical image classification using Convolutional Neural Networks (CNNs). As we embark on this project, we delve into the realm of leveraging cutting-edge deep learning techniques to augment traditional medical image analysis.

Probelm statement: You are give data which contain 100's of x-rays. Your task is to build a effective and efficient CNN model to classify them into Normal and Infected category. Show you amazing DL skills to bulid best model :)

Instructions

• For any concept/technique refer articles available on internet rather than using ChatGPT, as it may be misleading and many times provide only half information.
• Do not alter any pre-written code/comments.
• Write code in provided space only.
• Write commnets for what you did so that mentors can easy understand your work.
• Only use Google colab for running code.

Procedure

1. Fork and cloning this repository on your local device.
2. Open each task on Google colab.
3. Once task is completed download .ipynb file and store it in respective folder.
4. Name your file as Enrollment no.
   • For task1 store final file in Task1_solution with file name IIT2022119.
5. Push this file to forked repo and then send PR.
6. Your PR will be reviwed by the mentors. Once your PR is accepted, file will be merged and points will be granted.

Help

For any query feel free to contact [email protected]. You can also interact with mentors and community on Discord

medical-image-classification's Issues

Task2 - Image Pre-Processing

This task is open for all. No need to claim the issue.

In notebook Task2.ipynb first add code of Task1.ipynb done by you in previous issue as instructed.

Upload link of final colab file in folder *Task2_solutions * creating txt file with name as your Enrollment/roll number. Link should be in read mode and access to [email protected]. (Don't make file public).

Only complete task till plotting pre-processed images and send PR for this issue.

For resources refer Task2_Resources.txt

Lets deploy !!

Moment has come to deploy your model !!🎉

Use streamlit to deploy your DL model you have made so far. (You can use one of the two models that were made.)

May refer tutorials on youtube to learn how to deploy on streamlit platform.

How to submit:

• Create a folder of your name and in this add your code and a txt file with link to website.
• Example link: https://1st-deployment.streamlit.app/

Most beautiful website will be awarded maximum points.

Edit: Only best website will be awarded points. And there might be case, where no PR will be merged if your website is not found much attractive. Include graphics, animations, icons, fonts and many other ways to make it attractive.

Task2 - Image Augmentation

This task is open for all. No need to claim the issue.

Once your PR is merged for Pre-Processing task, start work on remaining portion of task i.e augmentation. Continue to work in same notebook you worked for Pre-processing..

Upload link of final colab file in folder *Task2_solutions * creating txt file with name as your Enrollment/roll number. Link should be in read mode and access to [email protected]. (Don't make file public). Refer Example.txt

For resources refer Task2_Resources.txt

Task 1

This task is open for all and every correct PR will be merged :)

Please attack link to colab file in read view while submitting PR

Note: Read Instructions_Task1.txt for resources and more details.

Task3 - Binary classification

The moment has come when you will train and test your model.

This is a competitive task and only the best model (based on ROC-AUC score and is model underfit or overfit). All other correct PR will be awarded 10 points. This issue is live till 20 dec, 23:59 !!

Make sure to setup GPU before start training.

Notebook: Task3.ipynb

Submit ur output in Task3_solutions. Inside this folder create a folder name and inside this add saved model ( .h5) and a txt file containing link of colab file.
Note: You have to submit both saved model and txt file

Edit:- Use only CNN (do not use pre-trained models). And while submitting output also submit all three graphs in jpeg format. And add test ROC-AUC score in txt file with file link.

Task4

Up to now you have trained your own CNN, now lets make predictions using pre-trained models.

This is a competitive task and only the best model (based on ROC-AUC score and is model underfit or overfit). All other correct PR will be awarded 10 points. This issue is live till 23 dec, 23:59 !!

Make sure to setup GPU before start training.

Notebook: Task4.ipynb

This is binary classification.

Submit ur output in Task4_solutions. Inside this folder create a folder name and inside this add saved model and a txt file containing link of colab file.

1. You have to submit both saved model and txt file
2. While submitting output also submit all three graphs in jpeg format.
3. Add test ROC-AUC score in txt file with file link.

Task2 - image preprocessing and augmentation